Professional Documents
Culture Documents
net/publication/228346619
CITATIONS READS
0 83
3 authors:
26 PUBLICATIONS 264 CITATIONS
University of Hartford
19 PUBLICATIONS 486 CITATIONS
SEE PROFILE
SEE PROFILE
D. Betsy Mccoach
University of Connecticut
146 PUBLICATIONS 5,472 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by D. Betsy Mccoach on 31 May 2014.
Jerome M. Katrichis
University of Hartford, USA
D. Betsy McCoach
University of Connecticut, USA
Abstract
Over the years, research on teacher bias towards females in mathematics has yielded
different results, which the study attributes to the narrow definition and measurement of
bias. This study validated the Females In Mathematics Scale (FIMS), to measure teacher
bias against females in mathematics. FIMS, based on the tripartite model of attitudes and
constructed using within-method triangulation, demonstrated acceptable levels of internal
reliability, and discriminant validity. Preliminary analyses showed that despite considerable
variability in what teachers think and how they feel about female students in mathematics,
they are more likely than not, to engage in unbiased behaviors in classroom interactions.
a categorical lens has a significant impact on the way bias is measured because it
distills the phenomenon to its most basic form (where it is either present or absent),
and as a result restricts the capabilities of what instruments can measure. We
believe that „mixed results‟ generated over years of research are largely a function of
narrow definition and measurement of bias, and we contend that the development
of robust instruments capable of capturing bias holistically hinges on how
adequately bias is operationalized and how well its nature as a continuous variable is
taken into account in instrument design or research methodology.
According to the Brophy and Good model (1974), teacher attitudes occur
along a continuum from „proactive‟ to „over reactive‟ with „proactive‟ teachers most
inclined to being unbiased and vice versa. Good (1987) contends that rather than
exist primarily at these extremes, teachers possess attitudes which vary in degrees
along the continuum, the majority of who hold light expectations which are
constantly adjusted as new information about the student emerges. Nevertheless,
subsequent research (Altermatt, Jovanovic and Perry, 1998; Cornbleth & Korth,
1980; Hartley, 1982; Tiedman, 2000) has focused mainly on investigating the
presence or absence of bias. To the best of our knowledge there have not been any
studies that have taken into account the varying degrees of its existence in its
measurement. Findings on bias should therefore be construed, not as „mixed‟ or
inconsistent, but as a testament to the nature of bias as a continuous variable and
the limitations of current instruments to capture this characteristic. So far results on
bias from the aforementioned studies point to individuals who lie at either end of
the spectrum, and in both cases, measures have not been designed to account for
the larger part of the story--one arguably central in reconciling both perspectives--
the mid-section where the majority of teachers lie. Incorporating this missing piece
of the puzzle in measurement requires that the operational definition of bias be
expanded to include all its dimensions in measurement (i.e. its cognitive, affect and
behavioral components).
Although the tripartite model defines attitudes as comprising cognitive,
affective and behavioral components (Olson & Zanna, 1993), most studies have
used one (Catsambis, 1995; Inzlicht & Ben-Zeev, 2003; Jones & Wheatley, 1990;
Sadker & Sadker, 1990) or sometimes two (Evertson, Brophy & Good, 1973; Good
& Brophy, 1972; Silberman, 1969), but not all three components in their research.
A multi-dimensional approach to the operationalization and measurement of bias
brings us closer to providing a more wholesome picture of the phenomenon. It also
provides an avenue to use within method methodological triangulation to
strengthen the validity of research findings. Triangulation uses multiple methods
and measurement procedures in research in order to increase validity (Ma &
Norwich, 2007). Methodological triangulation can be conducted using between
methods or within methods techniques where the former involves using contrasting
research methods to investigate a phenomenon e.g. using both a questionnaire and
observation in a single study (Denzin, 1970). In within methods triangulation,
varieties of the same method are used to investigate a research problem such as self-
report questionnaires with two contrasting scales to measure a construct (Jick,
1979). By providing cognitive, affective and behavioral bases upon which attitudes
can be evaluated; the tripartite model of attitudes enables the use of within methods
triangulation to not only investigate teacher bias along these components but also
strengthen the validity of these findings.
Here we briefly analyze methodological issues involved in the previous
measurement of teacher bias and attempt to remedy these issues by developing and
validating an instrument to measure teacher bias towards females in mathematics.
The validation of the Females in Mathematics Scale, FIMS, uses the tripartite
model of attitudes to expand the measured dimensions of bias, and within-method
triangulation to increase validity. Teacher bias is not only limited to females in math
and the sciences but also extends to males in subjects that are deemed
stereotypically „female‟ like reading (Hartley, 1982; Palardy, 1969), and to children
from lower socio-economic backgrounds (Alexander, Entwisle & Dauber, 1993;
Alvidrez & Weinstein, 1998). However, we chose to develop the Females in
Mathematics Scale (FIMS), because of the abundance of already existing literature
in the field on this topic, reports that teacher bias has a particularly profound
impact on performance and learning outcomes of girls in stereotypically male sex-
typed school subjects, such as math and science (Trouilloud et al, 2006), the
chronic scarcity of females pursuing careers in these domains (National Science
Foundation, 2000), and the fact that mathematics remains a critical component and
determinant of entry into careers in physical science domains like engineering.
Because teacher beliefs are filters that drive decisions regarding instruction (Fang,
1996) and have a significant impact on learning goals and outcomes of all students
(Eccles, 1993; Trouilloud, Sarrazin, Martinek, & Guillet, 2002), it is important that
it be correctly identified and reliably measured in order to design appropriate
interventions or enhance existing professional development programs.
Measurement and interventions could occur at any one of a number of levels from
state or district, to school or grade level. Additionally, once these interventions are
applied, a reliable, psychometrically sound instrument will also allow for the
measurement of attitude change and evaluating the effectiveness of such programs.
student interactions (Altermatt et al, 1998; Evertson, Brophy & Good, 1973; Good
& Brophy, 1972; Silberman, 1969) while self-report studies have relied mainly on
comparisons of teacher ratings of students‟ ability or achievement potential to
report teacher bias (Tiedmann, 2000; Alexander et al, 1993; Alvidrez & Weinstein,
1998).
under determined factors. They also found that approximately half of the published
applications reported used the wrong factor analytic technique (principal
components analysis vs. principal axis factoring) given their research goals. The
study by Cook and colleagues (2007) faced similar challenges; they used a sample
of 50 teachers to validate a four- item, four-factor scale (one item per factor)
measuring teachers‟ attitudes towards students with disabilities. Teachers rated a
total of 156 students with disabilities and 4 students without disabilities in each of
their classrooms on a 4-point Likert scale with responses ranging from not at all
true to extremely true for four statements: “I would like to keep this student for
another year for the sheer joy of it”; “I would like to devote all my attention to this
student because he or she concerns me”; “I would not be prepared to talk about
this student if his or her parents dropped by for a conference”, and lastly “If my
class was to be reduced, I would be relieved to have this student removed.” The
Pearson correlations for attachment, concern, indifference and rejection ratings are
quite good (values of .77, .70, .71 and .74 respectively) and compared to existing
measures, a considerable level of rigor was involved in the development of this
instrument. However, the authors used single item measures for each of the factors
and didn‟t use any of the conventional tools suggested for factor analysis to
determine factor structure of the items. As such, we have no way of knowing if their
items reflect the constructs being measured.
It appears that the development process of instruments to measure teacher
attitudes have been less rigorous than desired. Apparently measures of teacher bias
have been plagued with more serious problems than previously assumed:
unsatisfactory sampling procedures, relatively lax methodological procedure,
instruments with either unreported or weak psychometric properties, and restricted
range in measurement due to partial definition of bias. Earlier on, the use of
triangulation in measurement and research to improve reliability and validity of
findings was advocated, and the studies reviewed above make a clear case for that.
However, given the aforementioned problems instruments in research on bias
possess, it is unlikely that the use of triangulation with existing measures would
improve and add to the analysis. To do so would approximate pouring new wine
into old wineskins. Addressing teacher bias and remedying issues related to its
measurement would require that researchers return to the drawing board and
develop measures that: look at bias as a multi-dimensional construct, take into
account degree with which bias exists and extend its measurement beyond
dichotomy, follow stringent methodological procedure in research design, and
finally, subject instruments to rigorous validation to determine and strengthen
psychometric properties before use in research. This task, while daunting, is by no
means impossible and certainly carries with it rewards beneficial to strengthening
the overall quality of findings in research on teacher attitudes.
behavior), was used as a foundation for the creation of FIMS. The cognitive
component of attitude was defined as an individual‟s ideas, thoughts, perceptions,
beliefs, opinions or mental conceptualizations of the referent (Findler, Vilchinsky &
Werner, 2007); affect as the amount of positive or negative feelings one has towards
the referent, and behavior as one‟s intent or willingness to behave in a certain
manner toward the referent, or the actual behavioral response (Cook, 1992).
Method
The primary goal was to create items that would minimize response bias
and capture variability in responses along the seven point Likert scale, which ranged
from „strongly disagree‟ to „strongly agree‟. Items corresponding to each of these
three components of attitudes, specific to teaching were created and operationalized
as follows:
(1) Cognition: What teachers think about females‟ interest, effort and ability
in mathematics. High values on the seven- point scale for this factor would indicate
unbiased beliefs and a positive attitude about females in math and vice versa.
(2) Affect: Teachers‟ feelings about females‟ level of engagement, ability and
effort in their mathematics classes. Emotions from the pleasant-unpleasant axis of
the circumplex model of emotions (Russell & Barrett, 1999) were used to generate
items for affect. Teachers rated their feelings about negative levels of student
engagement, effort and performance in their mathematics classes. High scores
denote teachers‟ strong unpleasant feelings about poor performance or low
engagement and effort levels of females in mathematics, indicative of high
expectations and hence a more favorable attitude and vice versa.
(3) Behavior: Items that loaded under this factor concerned how teachers
respond to females in their mathematics classes. Items for this subscale were based
on Good‟s (1987) work regarding how teachers communicate expectations to
students and also on the Good and Brophy (1974) model. Items evaluated
differential teacher behavior in interactions between perceived high and low
achievers such that high values indicated biased behaviors and vice versa.
To summarize, unbiased teachers would have very high scores on cognition and
affect, and low scores on behavior.
Fifty items were created and used in a pilot study with 100 pre-service and
practicing math teachers. These items made direct comparisons between males and
females such as, “Females have a harder time focusing on math than males do.”
Participants also provided qualitative feedback on the items at the end of the
survey. Both qualitative feedback and results from the factor analysis were used in
tandem to assess the strength and uni-dimensionality of the items. Results signaled
weaknesses in the way items were worded.
Although the factor analysis yielded factors that denoted cognition, affect
and behavior, an extra factor emerged that contained a mix of cognitive and
behavior items. However, the cognition items had to do with „behavioral‟ indicators
of females in math e.g. Females don‟t put as much effort in math as males and
females have harder time focusing on math than males. These comparison
Content Validity
There were 33 items (sixteen cognition, seven behavior and ten affect) went
through three rounds of content reviews: a peer review, content validation by five
experts, and a final review by the investigators. The content experts checked either
„not relevant‟, „somewhat relevant‟, or „very relevant‟ to gauge the relevance of each
item as a measure for cognition, affect or behavior. A decision had been made a
priori to make 75% the appropriate cut off for correct item placement, with the
exception only for items that were believed to contribute conceptually to the
factors. Any items placed in the correct category by less than 75% of the experts
were deleted from the final instrument and Content Validity Indices, CVI‟s
(McKenzie, Wood, Kotecki, Clark & Brey, 1999) were computed based only on
the percentages in the „very relevant‟ category. But for one affect item, all items had
CVI‟s of 0.8 and above. Cognition and Behavior generated equally high scores
while „affect‟ had the lowest scores. Based on qualitative feedback from the experts
and scores on the CVI, three negatively worded items in the hypothesized „affect‟
factor were deleted, and replaced with positively worded items. The final number
of items remained the same.
Results
The FIMS KMO value of 0.863 suggested it was reasonable to run a factor
analysis on the data and Bartlett‟s test of sphericity was statistically significant (p
<.001). The eigenvalues above one reported seven factors and the total variance
explained by all seven factors was 52.91%. The scree plot (see Fig.1), parallel
analysis and MAP test, however, all suggested extracting three factors.
Figure 1
Scree plot
Scree Plot
10
8
Eigenvalue
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33
Factor Number
Analysis. Initially there were sixteen cognition, seven behavior and ten affect
items. After the EFA was run, the pattern matrix showed that 17 items loaded on
cognition, 10 on behavior and 5 on affect suggesting potential problems with cross
loadings among factors. In addition to this, a close examination of the factors in the
pattern matrix shown in table 1 revealed evidence of multi-dimensionality among
factors, discussed in detail below.
Table 1
Pattern Matrix for the Three-Factor Solution
Cont. Table 1
.71
q11. Most females find math enjoyable.
q14. Most females exert the effort required to succeed in
math .71
Cognition. All but two of the original cognition items loaded under the
hypothesized factor. Of the two items that failed to load on this factor, one loaded
under affect (q36) and the other (q24) did not load on any factors at all. Also, three
affect items cross- loaded very highly on cognition (q27 =.76, q37 =.73 and q41 =
.80). A possible explanation for this is that the wording tapped into both constructs
inadvertently contained a cognitive component. That is, they seemed to evaluate
one‟s thoughts about one‟s affect also (e.g. I am delighted by the performance of
females in math) toward females in math.
Affect. The affect subscale performed less optimally than cognition and
behavior subscales. Only four of ten items originally created for the scale loaded on
the factor, with seven others as already discussed, cross loading on cognition and
behavior. The items under affect were also multidimensional loading on both
affect and behavior (see table 2).
The cognition factor emerged strongest and explained the most variance
(27.5%) in teacher beliefs. Affect items performed relatively poorly compared to
the rest of the other items because of the way they were worded. Affect was
assessed relative to teachers‟ behaviors or thoughts about females in mathematics
and this seems to have been the underlying cause of multi-dimensionality or cross
loading in this case. Nevertheless, after excluding the cross loaded items, the
pattern and structure matrices for the three-factor solution respectively, also showed
moderate to high coefficients between the items and their respective factors.
Table 2
Structure Matrix for the Three-Factor Solution
Cont. Table 2
q13. Understanding math concepts comes naturally for .52
females
q40. Most females are capable of mastering advanced math
.52
topics with ease
q9. In general most females have an easy time learning math
.50
concepts.
q26. Females who excel in math are rare .39
q23. I tend to provide additional step by step support to
.72
females when solving simple math problems
q38. I give females additional support as they work through
.71
math problems
q16. I give females extra time to solve math problems .57
q19. I feel sorry for females who do not do well in math .55
q31. When females struggle with math problems, I'm quick to
.55
provide them with clues to the solution
q25. I encourage females to do more math practice outside
-.31 .44
class
q29. I pity females who do not succeed in math .44
q18. When correcting females in math, I try not to sound too
.43
negative
q33. I regret my efforts when females show no interest in
.39
math
q24. Usually, when a girl does poorly in math, it's because
she does not have the ability
q17. I am disappointed when females are bored in math class .38 .67
q32. It is upsetting when females neglect math .36 .62
q35. I get frustrated when my attempts to get females involved
.38 .60
in math do not work
q21. I take it personally when females in my class do not meet
.47 .50
achievement standards in math
q36. Usually, when a girl does poorly in math, it's because she
.41
hasn‟t put forth adequate effort
q12. I'm very direct when providing negative feedback to
females
Note. Extraction Method: Principal Axis Factoring. Rotation Method: Direct Oblimin with Kaiser
Normalization. Suppressed values below .30
Table 4
Communalities for the three- factor solution
Initial Extraction
q9. In general most females have an easy time learning math concepts .52 .29
q10. Most females are focused in math classes. .66 .44
q11. Most females find math enjoyable .71 .54
q12. I'm very direct when providing negative feedback to females .37 .08
q13. Understanding math concepts comes naturally for females .48 .31
q14. Most females exert the effort required to succeed in math .69 .56
q15. Most females are interested in math .66 .51
q16. I give females extra time to solve math problems .46 .37
q17. I am disappointed when females are bored in math class .57 .54
q18. When correcting females in math, I try not to sound too negative .43 .21
q19. I feel sorry for females who do not do well in math .44 .34
q20. Most females do well in math .63 .54
q21. I take it personally when females in my class do not meet achievement
.50 .40
standards in math
q22. Most females like taking math classes .74 .66
q23. I tend to provide additional step by step support to females when solving
.55 .55
simple math problems
Cont. Table 4
q24. Usually, when a girl does poorly in math, it's because she does not have
.20 .03
the ability
q25. I encourage females to do more math practice outside class .42 .32
q26. Females who excel in math are rare .49 .33
q27. I am happy with the level of interest most females show in math .66 .55
q28. Most females work hard at math .75 .57
q29. I pity females who do not succeed in math .40 .23
q30. Most females are engaged in math class .72 .68
q31. When females struggle with math problems, I'm quick to provide them
.42 .29
with clues to the solution
q32. It is upsetting when females neglect math .47 .46
q33. I regret my efforts when females show no interest in math .43 .23
q34. Most females spend the time necessary to master mathematics concepts .70 .59
q35. I get frustrated when my attempts to get females involved in math do not
.48 .48
work
q36. Usually, when a girl does poorly in math, it's because she hasn‟t put
.39 .24
forth adequate effort
q37. I am delighted with the performance of most females in math .73 .64
q38. I give females additional support as they work through math problems .54 .51
q39. Most females are motivated in math .80 .79
q40. Most females are capable of mastering advanced math topics with ease .54 .31
q41. I am pleased by the level of effort most females devote to math .78 .68
Reliability analysis. Based on the results from the factor analysis, items to
conduct a reliability analysis were selected. An item was excluded from the
reliability analysis if it did not load on its hypothesized factor, if it loaded on its
factor but with a value less than .33 and if it had communalities less than .35
coupled with low loading on its factor. Acceptable (moderate) communality
coefficients range between .4 and .7 (Fabrigar et al., 1999) so the above range was
used as a baseline to retain items with communalities of .35 and above. Twenty-
three items (fourteen cognition, six behavior and three affect), rated on a seven-
point Likert scale were used to run the reliability analysis in SPSS. Corrected item
total correlations, squared multiple correlations, average inter item correlations
(IIC), standard deviations, the inter-item correlations matrix, a desirable alpha level
of 0.8 set a priori, and Cronbach‟s alpha if item is deleted informed decisions about
the reliability analysis of the subscales.
Table 5
Summary of reliability analysis for FIMS subscales
Affect. The affect subscale consisted of three items, with a Cronbach‟s alpha
of 0.75. The corrected item-total correlations were moderate (.57 - .60) and the
same was true for squared multiple correlations that had values ranging from .33 to
.36.
Behavior. The behavior subscale had a reliability coefficient of .73, and
reported moderate item total correlations among its items (.33 -.62). As was the
case with the affect subscale, alpha if deleted did not provide any items that would
increase reliability of the scale if deleted so no items were removed.
All 23 items used to conduct the reliability analysis were retained. Based on
calculations using the Spearman-Brown Prophecy formula, it was determined that
the affect scale would need one additional item to bring its alpha to .8 and behavior
would need to have three additional items to obtain an alpha of .8.
Discussion
So, are teachers biased towards females in mathematics? And if so, to what
extent? The development of FIMS was inspired by two gaps in the literature on
teacher attitudes: the general lack of psychometrically sound instruments to
measure bias as multi-dimensional construct and the limited ability of current
measures to report bias in gradations.
Table 6
Composite sub-scale scores by group compared against FIMS composite score
Cognition:
4.51 4.72 4.90 4.81 4.50 4.65
Mean 0.71 0.82 0.85 0.83 0.71 0.79
SD
Affect:
4.43 4.23 4.14 4.19 4.38 4.29
Mean 1.08 1.52 1.65 1.42 1.27 1.39
SD
Behavior:
4.77 5.06 4.90 4.83 5.11 4.96
Mean 0.97 1.05 1.06 1.15 0.94 1.03
SD
Results also show little difference in means for cognition and behavior
between male and female teachers or even among teachers in elementary, middle
or high school. Although mean scores for affect were generally lower than for
cognition and behavior (M = 4.29, SD = 1.39 vs. M = 4.65, SD = .79 and M = 4.96,
SD = 1.03, respectively), group means for affect varied more compared to the
group means for cognition and behavior. A comparison of means across gender
reveals higher affect means for male teachers (M = 4.43, SD = 1.08 vs. females M =
4.23, SD = 1.52). Similar comparisons across grade levels also indicate higher affect
means high school teachers (M = 4.38, SD = 1.27) compared to middle school (M
= 4.19, SD = 1.42) and elementary school teachers (M = 4.14, SD = 1.65). These
results suggest that compared to their respective reference groups (gender and
grade level) males (and high school teachers) possess slightly more positive than
neutral feelings towards their students. The differences are not significant though,
since all groups generally were closer to neutral than slightly favorable in their
evaluative judgments, feelings and behaviors towards female students in
mathematics. These results bolster the theory that teachers attitudes are neutral,
and that over time most teachers adjust their expectations accordingly as more
information about the student becomes available (Good, 1987).
Table 7
Subscale correlations
Cognition
Pearson Correlation __ -.11 .01
Sig. (2-tailed) .14 .94
Affect
Pearson Correlation __ .31**
Sig. (2-tailed) .00
Behavior
Pearson Correlation __
Sig. (2-tailed) .
expectations toward these students, all behaviors which have been previously linked
with students that teachers like (Good & Brophy, 1972) and also with males as
opposed to females (Sadker & Sadker, 1989).
Degree of bias. The FIMS scale is still in its initial stages of development
and as such, cannot offer individual assessment of degree of bias. It does, however,
provide insight on general trends of degree by group. Analyses of means and
standard deviations show that even though most teacher attitudes towards females
in math generally tend to be neutral, there is considerable variability in beliefs,
affect and behavior among teachers. Of notable significance here is the spread in
the measures, especially for the affect subscale.
Standard deviations for the cognition subscale were smaller than those of
affect or behavior, and within close range (SD = 0.71- 0.85) for males, females and
different grade levels. Standard deviations for behavior were equally homogenous
(SD = 0.94 - 1.15) across all groups. On the other hand, the standard deviations for
affect revealed a higher level of heterogeneity among the groups sampled with
values ranging from 1.08 to 1.65 compared to the other subscales. This seems to
suggest that while teachers have similar beliefs about female students there is great
variability with respect to how they feel about the abilities, effort and level of interest
these students show in mathematics, especially among elementary school teachers
(SD = 1.65) and female teachers (SD = 1.52) compared to other grade levels, and
males, respectively.
The development of affect items for the validation of FIMS was a challenge.
A number of affect items were multidimensional, loading on other factors. This
greatly reduced the number of affect and behavior items that were used in the
reliability analysis. The scale could have benefited from more measures for these
constructs. Additionally, the scale has only undergone an exploratory factor analysis
and is in its early stages of development. Findings presented here cannot and
should not be generalized to the teacher population before confirmatory factor
analyses on different samples have been conducted.
Finally, although FIMS was developed to evaluate teacher bias towards
students in mathematics, it is not mathematics specific. It can and should be
adapted to evaluate teacher beliefs about students belonging to groups in other
disciplines for which negative stereotypes apply, after it has been formalized.
Conclusion
References
Albarracín, D., Johnson, B. T., & Zanna, M. P. (Eds.) (2005). The handbook of
attitudes. Mahwah, NJ: Lawrence Erlbaum.
Alexander, K. L., Entwisle, D. R., & Dauber, S. L. (1993). First grade classroom
behavior: Its short and long-term consequences for school performance.
Child Development, 64, 801-814.
Altermatt, R. E., Jovanovic, J., & Perry, M. (1998). Bias or responsivity? Sex and
achievement-level effects on teachers‟ classroom questioning practices.
Journal of Educational Psychology, 90, 516-527.
Alvidrez, J. & Weinstein, R. S. (1999). Early teacher perceptions and later student
academic achievement. Journal of Educational Psychology, 91, 731-741.
American Association of University Women. (2010). Why so few?: Women in
science, technology, engineering and mathematics. (Research Report. No.
077-10 5M). Washington, DC. Hill, C., Corbett, C., & St. Rose, A.
Antonak, R. F., & Livneh, H. (2000). Measurement of attitudes towards persons
with disabilities. Disability and Rehabilitation, 22, 211–224.
Brophy, J. (1987). Synthesis of research on strategies for motivating students to
learn. Educational Leadership, 45(2), 40-48.
Brophy, J., & Good, T. (1974). Teacher-student relationships: Causes and
consequences. New York: Holt, Rinehart & Winston.
Catsambis, S. (1995). Gender, race, ethnicity and science educations in the middle
grades. Journal of Research in Science Teaching, 32, 243-257.
Cook, G. B., Cameron, L. D., & Tankersley, M. (2007). Inclusive teachers‟
attitudinal ratings of their students with disabilities. The Journal of Special
Education, 40, 230–238.
Cook, D. (1992). Psychological impact of disability. In R. M. Parker & E. M.
Szymanski (Eds.), Rehabilitation counseling basics and beyond (pp. 249–
272). Austin, TX: PRO-ED.
Cornbleth, C., & Korth, W. (1980). Teacher perceptions and teacher-student
interaction in integrated classrooms. Journal of Experimental Education, 48,
259-263.
Denzin, N. K. (1970). The research act in sociology. Chicago: Aldine.
Eccles, J. S. (1993). School and family effects on the ontogeny of children‟s
interests, self perceptions, and activity choice. In J. Jacobs (Ed.), Nebraska
Symposium on Motivation: Vol. 40. Developmental perspectives on
motivation (pp. 145-208). Lincoln, NE: University of Nebraska Press.
Evertson, C., Brophy, J., & Good, T. (1973). Communication of teacher
expectations: Second grade. (Report 92). The University of Texas at Austin,
Austin: Research and Development Center for Teacher Education.
Fabrigar, R. L., Wegener, T.D., MacCallum, C. R., & Strahan, J. E. (1999).
Evaluating the use of exploratory factor analysis in psychological research.
Psychological Methods, 4(3), 272-299.
McKenzie, J. F., Wood, L. M., Kotecki, E. J., Clark, K. J., & Brey, A. R. (1999).
Establishing content validity: Using qualitative and quantitative Steps.
American Journal of Health and Behavior, 23(4), 311-318.
Murphy, M. C., Steele, C. M., & Gross, J. J. (2007). Signaling threat: How
situational cues affect women in math, science and engineering settings.
Psychological Science, 18(10), 879- 885.
National Science Foundation. (2000). Women, minorities, and persons with
disabilities in science and engineering (NSF Rep. No. 00-327). Arlington,
VA: Bleeker, M. M.
Olson, J. M., & Zanna, M. P. (1993). Attitudes and attitude change. Annual Review
of Psychology, 44, 117-154.
Palardy, J. M. (1969). What teachers believe- what children achieve. Elementary
School Journal, 69, 370-374.
Pett, M. A., Lackey, R. N., & Sullivan, J. J. (2003). Making sense of factor analysis:
The use of factor analysis for instrument development in healthcare
research. London, United Kingdom: Sage.
Rosenfeld, M., & Rosenfeld, S. (2008). Developing effective teacher beliefs about
learners: the role of sensitizing teachers to individual learning differences.
Educational Psychology, 28, 245-272.
Russell, J. A., & Barrett, L. F. (1999). Core affect, prototypical emotional episodes,
and other things called emotion: Dissecting the elephant. Journal of
Personality and Social Psychology, 76, 805-819.
Sadker, M., & Sadker, D. (1990). Confronting sexism in the college classroom. In
S. K. Gabriel & I. Smithson (Eds.), Gender in the classroom: power and
pedagogy (pp. 176-187). Urbana: University of Illinois Press.
Silberman, M. L. (1969). Behavioral expression of teachers‟ attitudes toward
elementary school students. Journal of Educational Psychology, 60, 402-
407.
Stevens, J. (1996). Applied multivariate statistics for the social sciences (3rd ed.).
Mahwah, NJ: Lawrence Erlbaum.
Tiedemann, J. (2000). Parents‟ gender stereotypes and teachers‟ beliefs as
predictors of children‟s concept of their mathematical ability in elementary
school. Journal of Educational Psychology, 92(1),144-151.
Trouilloud, D., Sarrazin, P., Bressoux, P., & Bois, J. (2006). Relation between
teachers‟ early expectations and students‟ later perceived competence in
physical education classes: Autonomy supportive climate as a moderator.
Journal of Educational Psychology, 98(1), 75-86.
Trouilloud, D., Sarrazin, P., Martinek, T. G., & Guillet, E. (2002). The influence
of teacher expectations on students‟ achievement in physical education
classes: Pygmalion Revisited. European Journal of Social Psychology, 32(5),
591-607.
Velicer, W. F., & Flava, J. L. (1998). Effects of variable and subject sampling as
factor pattern recovery. Psychological Methods, 3, 231-251.